A Fixed Basket, Priced Twice, Thirty-One Years Apart
Credit: Daily Vitality
In July 1995, a shopper could walk out of an American supermarket with four pounds of ground beef, two gallons of whole milk, two dozen eggs, a whole chicken, a pound of bacon, a pound of cheddar, five pounds of potatoes, four pounds of bananas, a pound of ground coffee and eleven other staples, hand the cashier a fifty-dollar bill, and get change. Not roughly fifty dollars. Fifty dollars and zero cents.
That exact cart, priced item by item from the same federal price series in July 2026, costs $123.26.
The obvious conclusion is that groceries have outrun inflation. The obvious conclusion is wrong — or rather, it is wrong for nineteen of the twenty things in the cart.
Inflation coverage reports one number. One number is a terrible instrument for a question like this, because a grocery bill is not one thing; it is twenty things moving in different directions at different speeds, and the average conceals every interesting fact about them. So this article does something the monthly inflation story cannot: it fixes a basket, freezes the quantities, prices it at two dates thirty-one years apart, and ranks every item by how far it broke away from the average.
What comes out is not the article anyone expects. The basket did outpace inflation. It did so almost entirely because of one item. And the middle of the distribution — the median line in the cart — did not move at all.
The instrument is deliberately simple, because a simple instrument is one a reader can check.
Every price here comes from the U.S. Bureau of Labor Statistics Average Price Data program, which publishes the actual per-unit retail prices its field staff collect for the Consumer Price Index — dollars per pound, per dozen, per gallon, U.S. city average, every month. These are not index points. They are prices.
The basket holds twenty of them. Quantities were chosen before any pricing was done, to resemble a plausible weekly shop for a small household, and then frozen. Nothing substitutes, nothing gets dropped, nothing gets swapped for the store brand. The July 1995 total landing on exactly $50.00 was luck, not design, and it is the kind of luck worth admitting to rather than dressing up.
Three method choices carry most of the weight.
Same calendar month, both times. The comparison runs July 1995 to July 2026 — thirty-one years to the month. Holding the month constant strips out seasonality, which matters enormously for lettuce, tomatoes and citrus, and matters more than most people realize for eggs.
July rather than January. The per-gallon whole-milk series begins in July 1995; before that, BLS priced milk by the half gallon on a series it retired in 1997. Starting in July lets milk into the basket without stitching two different series together.
No splices, at all. Every one of the twenty series runs continuously from July 1995 to July 2026 under a single item code and a single unit definition. Where a comparison would have required chaining across a redefinition, the item was excluded rather than patched. That decision cost the basket some obvious groceries, and later in this piece we name every one of them.
Real changes are calculated by deflating each nominal change by the all-items CPI-U, which rose from 152.5 to 333.918 across the span — a multiplier of 2.1896. In plain terms: general inflation would have turned $50.00 into $109.48.
Throughout this article, "nominal" means the change in sticker price and "real" means the change after general inflation is removed. An item that rose 119% nominally over these 31 years did not get more expensive in real terms — it simply kept pace.
Our companion piece on the real cost of eating healthy in every state runs the same kind of fixed-basket logic across geography rather than time; this one is the time axis of the same question.
Here is the headline, and here is the complication that comes attached to it.
The basket rose from $50.00 to $123.26 — a multiplier of 2.465, or +146.5% in nominal terms. Against a CPI-U multiplier of 2.1896, that is a real increase of 12.6%. Expressed the other way around: the 2026 cart costs $56.29 in 1995 dollars, against the $50.00 that was actually paid. A $50 bill covers 40.6% of it today.
Now the complication. The official Consumer Price Index for food at home rose from 148.2 to 321.643 over the same thirty-one years — a multiplier of 2.170, which is below the all-items figure. By the government's own headline measure, groceries got about 0.9% cheaper in real terms since 1995. In the most recent reading, published August 12, 2026, the Bureau of Labor Statistics reported all-items inflation of 3.4% over twelve months and food-at-home inflation of 2.7%.
So which is it? Both, and the gap between them is the most useful thing in this article.
A fixed basket allows no substitution. It buys four pounds of ground beef in 2026 because it bought four pounds in 1995, no matter what has happened to the price. The CPI does not work that way: it reweights as spending patterns shift, and it adjusts for quality change. Real shoppers behave like the index, not like the basket — they see $6.885 a pound and buy something else.
That is the whole story of the 0.295 difference between 2.465 and 2.170. It is the measurable cost of refusing to substitute. And it is why the grocery bill can feel worse than the index says while the index remains perfectly accurate: the index is describing what people ended up buying, and the shopper is remembering what they used to buy.
Every line below is a real BLS series with a published item code. The unit prices are as published, unrounded. The series ID column is there on purpose: anyone with a spreadsheet and the BLS average price flat file can rebuild the $123.26 exactly.
Totals are computed from unrounded prices; line items are rounded to the cent, so the 1995 column visibly sums to $50.01.
Read down the "Real" column and the actual finding appears. Nine items beat inflation. Nine lost to it. Two — flour and ice cream — landed so close to it that they matched to within a tenth of a percentage point.
The median item changed 0.0% in real terms. In thirty-one years, the middle of an American grocery cart went precisely nowhere.
What moved was the spread. In 1995 the cart's most and least inflationary items were separated by a wide but ordinary gap. By 2026 the top of the table sits at +130% real and the bottom at −43%, and the same twenty foods that once behaved roughly alike now behave like assets in different markets.
One line in that table is doing an extraordinary amount of work.
Four pounds of ground beef cost $5.46 in July 1995 and $27.54 in July 2026. That single item accounts for 30.1% of the entire dollar increase in the basket, and its share of the total bill went from 10.9% to 22.3% — the same four pounds, now consuming nearly a quarter of the cart.
Remove it, and the finding inverts. The remaining nineteen items went from $44.54 to $95.72, a multiplier of 2.149 — that is 1.9% below general inflation, and statistically indistinguishable from the official food-at-home index. The headline of this article survives or dies on one product.
The tempting story is that beef exploded after 2020. The data does not support it. Indexed to July 1995 and adjusted for inflation, ground beef sat at 101.8 in 2000, 131.4 in 2005, 125.7 in 2010, 196.6 in 2015, 183.9 in 2020 and 230.4 in 2026.
Read that sequence again. Real beef prices were flat through 2000, stepped up in the mid-2000s, climbed hard through the mid-2010s, and then accelerated again after 2020. The 1995-to-2020 stretch alone delivered a real increase of 83.9%; the six years since added 25.3%. Roughly a quarter of the total gain is recent. Three-quarters of it accumulated while nobody was writing headlines about it.
The supply side explains the recent leg. The American Farm Bureau Federation reports the national cattle herd at 86.2 million head as of January 1, 2026 — down 300,000 head year over year and the smallest in roughly seventy-five years, a level last seen around 1951. Multi-year drought reduced pasture and delayed rebuilding. New World screwworm detections restricted feeder-cattle imports from Mexico. And a cattle herd cannot be rebuilt quickly; the biology sets the schedule, not the market.
The bottom of the table is the half of the story nobody tells.
Bananas rose 24.5% in nominal terms across thirty-one years — from $0.522 to $0.650 a pound. Deflated, that is a 43.1% collapse. A banana is one of the very few things in American life that is meaningfully cheaper than it was when Toy Story was in theaters. Trade liberalization, container-shipping efficiency and plantation yield gains did that, and they did it quietly.
Dairy is the more surprising entry. Whole milk fell 20.5% in real terms and cheddar 23.3%, on the back of decades of productivity gains per cow and consolidation in processing. Dry pasta fell 26.0%. Field tomatoes fell 26.9%.
Eggs deserve a separate note, because they are the most emotionally available price in the country. The national average hit $6.227 per dozen in March 2025 and stood at $2.189 in July 2026 — a figure you can read straight off the BLS chart package that accompanies each CPI release. Across the full thirty-one years, eggs are up just 13.7% in real terms. They are volatile, not structurally expensive, and a single-month snapshot of them will mislead you in either direction.
One caution belongs here. "Cheaper" in this data can also mean "different," because average prices are not quality-adjusted — a topic we take apart in our look at 1970s supermarket staples versus their 2020s reformulations.
Prices are one unit of account. For a cart, protein is another — and converting between them changes what the table looks like.
Using protein values from USDA FoodData Central, ground beef at 17.2 grams of protein per 100 grams carries about 78 grams in a pound as purchased. Run every price through its own conversion and the cost of 30 grams of protein looks like this:
In July 2026, the cheapest 30 grams of protein in this cart costs about seventy cents, and it is pasta. The most expensive costs $2.65, and it is ground beef. Across these five foods, the gap between the cheapest and priciest source widened from 2.8 times in 1995 to 3.8 times in 2026.
A word on what this table is and is not. Because protein content per unit is held constant, the conversion cannot re-rank anything — it is the price ranking wearing different units. Its value is in the levels: it shows how far apart the options drifted, which is a different observation from "protein got expensive." Protein did not get expensive. Some protein did. For a fuller treatment of that specific question, we ranked the best protein per dollar across 25 foods separately.
There is a third unit of account, and it produces the finding that most complicates everything above.
Average hourly earnings for production and nonsupervisory workers — the long-running BLS series carried on FRED — went from $11.67 in July 1995 to $32.40 in July 2026. That is a multiplier of 2.776, against the basket's 2.465. Wages, on this measure, outran the cart.
Converted into time: the basket cost 4.28 hours of work in 1995 and 3.80 hours in 2026 — about 11% less.
This finding is narrow and should be read narrowly. It uses gross pre-tax earnings from one wage series covering production and nonsupervisory workers, and it says nothing whatsoever about housing, healthcare, childcare or education — which is where household budget pressure actually concentrates. The correct reading is "this specific cart got cheaper in work-hours," not "people are better off."
We report it because leaving it out would make the article more persuasive and less true.
Two gaps sit inside this dataset, and showing them is more useful than smoothing over them.
October 2025 does not exist. Fifty-six of the fifty-nine national average-price series that were active on either side of that month have no value for it. The cause was a lapse in appropriations, and the Bureau of Labor Statistics states plainly that it could not collect October 2025 reference period survey data for some surveys, "resulting in missing data values where data could not be retroactively collected." The October 2025 CPI release was canceled outright; the November release was pushed to December 18, 2025. The gap is permanent — the observations were never made, and cannot be recovered. On BLS's own published tables, September 2025 is followed directly by November 2025.
Spring and summer 2020. Collection disruption left holes across produce series between March and September of that year. Within our twenty, iceberg lettuce is missing 42 months since July 1995, coffee 41 and rice 26. Every other item in the basket is missing exactly one month, and that month is October 2025.
Neither gap touches July 1995 or July 2026, and no value in the basket carries a BLS footnote code. Nothing here was interpolated, smoothed or filled. Where the data is absent, it is shown as absent — including, importantly, in 2026 reference months, which were unaffected because collection resumed normally.
Six limitations, stated before anyone else states them.
These are national averages, not any real store. Regional pricing, chain pricing, warehouse clubs and store brands all diverge from them, sometimes sharply. That divergence is its own subject, which we take up in our comparison of which grocery stores offer the best nutrition per dollar.
Average prices are not the same instrument as the CPI. The BLS fact sheet on average prices is explicit that they estimate price levels while the index measures price change, and it warns that roughly a quarter of outlets and products rotate through the sample each year — a shift from a pricier outlet to a cheaper one moves the average even when no price moved.
Some item definitions are narrow, which makes some samples thin. In BLS's phrasing, the number of prices behind an average "can be substantially smaller than the number of prices used in index calculation." Coffee, lettuce and rice are the three lines in our table where that caution bites hardest, and their missing-month counts show why.
Package size is normalized, product quality is not. Because the series are per pound, per dozen and per gallon, headline shrinkflation is already stripped out. What is inside the pound is not held constant. A 2026 pound of ground beef and a 1995 pound are not guaranteed to be the same product.
No substitution. Discussed above, and worth restating: this is the single largest reason the basket diverges from the official index.
Two months, not two eras. To test whether the result is an artifact of picking convenient months, we reran the entire basket on twelve-month averages — August 1995 through July 1996 against August 2025 through July 2026. That produces $50.43 rising to $123.93, a multiplier of 2.458 against the single-month 2.465. The two methods agree to within seven thousandths.
The no-splice rule has a visible cost, and it is worth naming, because the absences are conspicuous.
BLS has retired the price series for a long list of 1995 staples. Butter was last priced in April 2012 and resumed under a new code in April 2018, leaving a six-year hole. Peanut butter ended in December 2017. Red Delicious apples ended in October 2017, whole frozen turkey in February 2020, canned light tuna in September 2017, bologna in September 2019, frankfurters in April 2022, Thompson Seedless grapes in September 2021, carrots in March 2014, cabbage and celery in October 2012, and broccoli in March 2020.
Items leave the list when they stop being purchased commonly enough to price reliably, and a few return years later under a new item code and a changed definition — which is precisely what a same-item comparison cannot survive. The national grocery basket is not a fixed object. It is a moving target, revised as the country's eating habits revise themselves.
The instinct behind "everything used to be cheaper" is sound. The arithmetic is not.
Over thirty-one years, an American grocery cart tracked general inflation almost exactly — until you look inside it, at which point the tidy average dissolves into twenty items that agree about nothing. One protein source more than doubled in real terms. One fruit lost nearly half its real price. Half the cart beat inflation, half of it lost, and the item sitting at the median did not move at all.
That is the finding, and it is more interesting than the complaint it replaces. Prices did not rise together. They scattered — and the scatter, not the average, is what a shopper actually experiences at the register.
The basket is a fixed definition, and we intend to price it again each July. Readers wanting to see the same cost-per-nutrient logic applied to a working meal plan can find it in our breakdown of whether 100 grams of protein a day is achievable on $5.
About the Author
Ryan May is a digital publishing leader with more than 25 years of experience turning complex information into clear, useful editorial work. At Daily Vitality, he focuses on evidence-led reporting about nutrition, fitness, healthy aging, and long-term wellbeing, informed by his own interests in cooking, yoga, cycling, and sustainable healthy habits.
Because a fixed basket allows no substitution. It buys four pounds of ground beef in 2026 because it bought four pounds in 1995, regardless of price. The Consumer Price Index reweights as spending patterns change and adjusts for quality. The basket rose 12.6% in real terms; the food-at-home index fell about 0.9%. Both are correct measurements of different questions.
Holding the calendar month constant removes seasonality, which distorts produce and citrus badly and affects eggs more than people expect. July 1995 to July 2026 is also exactly 31 years. As a check, the same basket run on 12-month averages gives a multiplier of 2.458 against the single-month 2.465 — close enough that the choice of month is not driving the result.
They come from the BLS Average Price Data program, which publishes actual per-unit retail prices collected for the CPI. Every item in the table carries its series ID, and the underlying flat file is public. With a spreadsheet, the quantities listed and those series IDs, the $123.26 total is fully reproducible.
Because BLS retired those series. Butter was last priced in April 2012 and only resumed under a new item code in April 2018, leaving a six-year gap; peanut butter ended in December 2017 and whole frozen turkey in February 2020. Including them would have required splicing across a definition change, and this basket uses no splices.
Partly. The series are priced per pound, per dozen and per gallon, so a smaller package at the same price shows up as a higher unit price — headline shrinkflation is already normalized out. What is not held constant is the quality or composition of what is inside that pound.
A lapse in appropriations meant the data was never collected, and BLS states it could not be gathered retroactively. Fifty-six of the fifty-nine active national average-price series have no October 2025 value, and the October 2025 CPI release was canceled outright. The gap is permanent, and it does not touch either month used in this comparison.
On this specific measure, yes. Average hourly earnings for production and nonsupervisory workers rose faster than the basket — a multiplier of 2.776 against 2.465 — so the cart cost about 11% fewer work-hours in 2026 than in 1995. That figure uses gross pre-tax earnings from one wage series and says nothing about housing, healthcare, childcare or education.
Supply. The national cattle herd stood at 86.2 million head on January 1, 2026, roughly a 75-year low, after multi-year drought reduced pasture and delayed rebuilding, and screwworm detections restricted feeder-cattle imports from Mexico. Rebuilding a herd takes years for biological reasons. Notably, most of beef's real increase predates 2020 rather than following it.
Almost certainly not exactly. These are U.S. city averages. Regional markets, individual chains, warehouse clubs and store brands all price differently, and BLS itself cautions that sample rotation between outlets moves the average even when no underlying price has changed.
